TRAC #14504
Signed-off-by: Daniel Koch

Author:    Nicolas Capens

git-svn-id: https://angleproject.googlecode.com/svn/trunk@494 736b8ea6-26fd-11df-bfd4-992fa37f6226

TRAC #14506
This speeds up surface swapping by avoiding an unnecessary StretchRect from a render target surface to a render target texture (anti-aliased surfaces are unsupported so no resolve is needed).
Signed-off-by: Daniel Koch

Author:    Nicolas Capens

git-svn-id: https://angleproject.googlecode.com/svn/trunk@493 736b8ea6-26fd-11df-bfd4-992fa37f6226

TRAC #14503
Signed-off-by: Daniel Koch

Author:    Nicolas Capens

git-svn-id: https://angleproject.googlecode.com/svn/trunk@492 736b8ea6-26fd-11df-bfd4-992fa37f6226

Implemented validation for loop and indexing limitations specified by GLSL ES spec 1.0 Appendix A Section 4 and 5.
A couple of things to note:
- This CL only validates the "form" of loop and indexing. It does not detect number-of-iterations or out-of-bound access. This will require more involved analysis/heuristics.
- I haved combined SH_VALIDATE_CONTROL_FLOW and SH_VALIDATE_INDEXING into one flag - SH_VALIDATE_LOOP_INDEXING. Validating both together is much easier.
BUG=48
Review URL: http://codereview.appspot.com/3225041

git-svn-id: https://angleproject.googlecode.com/svn/trunk@491 736b8ea6-26fd-11df-bfd4-992fa37f6226

For textures that are not already render targets, use the CPU to filter them and mark them as dirty.

I tested it by modifying the "mip_map_2d" demo to call glGenerateMipmap instead of generating the mipmaps itself. I also used the debugger to force it to take the render target path and verified that it still worked.

We've been having problems with the display being reset when video memory pressure is heavy with one of our demos. This patch made a noticable improvement on Vista but I don't think it will necessarily help on XP.

Review URL: http://codereview.appspot.com/3262041

git-svn-id: https://angleproject.googlecode.com/svn/trunk@490 736b8ea6-26fd-11df-bfd4-992fa37f6226

Moved the global-pool-allocator to TCompiler so that all memory allocated by the compiler can be de-allocated. Earlier the global-pool-allocator kept accumulating memory from all compilers (symbol-table in particular). The memory was only de-allocated when gpu-process exited or ShFinalize() was called. This was a problem for Chromium which keeps the GPU process around for the browser session. Now the memory is de-allocated as soon as the compiler is deleted, which happens when a tab is closed.
BUG=58808 (crbug.com)
Review URL: http://codereview.appspot.com/3280041

git-svn-id: https://angleproject.googlecode.com/svn/trunk@489 736b8ea6-26fd-11df-bfd4-992fa37f6226

TRAC #14429
Signed-off-by: Daniel Koch

Author:    Nicolas Capens

git-svn-id: https://angleproject.googlecode.com/svn/trunk@488 736b8ea6-26fd-11df-bfd4-992fa37f6226

TRAC #13625
This patch ensures that (member variables of) arrays of structures are listed separately by glGetActiveUniforms(). Their location can also be retrieved individually with glGetUniformLocation() using the array operator.
Signed-off-by: Daniel Koch

Author:    Nicolas Capens

git-svn-id: https://angleproject.googlecode.com/svn/trunk@487 736b8ea6-26fd-11df-bfd4-992fa37f6226

Resize surface on receipt of WM_SIZE to avoid corruption during resize. We hook WM_SIZE using window subclassing.

This is a continuation of http://codereview.appspot.com/3038042/


Review URL: http://codereview.appspot.com/3122041

git-svn-id: https://angleproject.googlecode.com/svn/trunk@486 736b8ea6-26fd-11df-bfd4-992fa37f6226

Removed actions for generating parser-lexer now that they have been checked in. Also removed hooks to pull cygwin - it was only required for flex and bison that are not required anymore. If a developer needs to modify grammar files, he is expected to install cygwin.
Review URL: http://codereview.appspot.com/3177042

git-svn-id: https://angleproject.googlecode.com/svn/trunk@485 736b8ea6-26fd-11df-bfd4-992fa37f6226

git-svn-id: https://angleproject.googlecode.com/svn/trunk@484 736b8ea6-26fd-11df-bfd4-992fa37f6226

TRAC #14390
Signed-off-by: Daniel Koch

git-svn-id: https://angleproject.googlecode.com/svn/trunk@483 736b8ea6-26fd-11df-bfd4-992fa37f6226

Replaced TIntermLoop::testFirst with TIntermLoop::loopType to clearly indicate which type of loop it is. In some cases it is not possble to differentiate between a for-loop and while-loop.
BUG=48
Review URL: http://codereview.appspot.com/3123041

git-svn-id: https://angleproject.googlecode.com/svn/trunk@482 736b8ea6-26fd-11df-bfd4-992fa37f6226

I suspect there is a subtle reason why this is the wrong thing to do.

The issue I am trying to address is glBlitFramebufferANGLE generating an error when bliting from a GL_RGB8_OES renderbuffer to a GL_RGB texture bound to the read framebuffer and draw framebuffer respectively. It verifies that both surfaces have the same underlying D3DFMT but GL_RGB8_OES renderbuffers are backed by X8R8G8B8 and GL_RGB textures were previously backed by A8R8G8B8.

Since textures cannot as far as I know be created GL_RGB8_OES, it would be useful to have a way to blit from a GL_RGB8_OES renderbuffer to a GL_RGB texture.




Review URL: http://codereview.appspot.com/2822041

git-svn-id: https://angleproject.googlecode.com/svn/trunk@481 736b8ea6-26fd-11df-bfd4-992fa37f6226

I've been seeing crashes like this on Windows XP:

0x013319aa 	[libglesv2.dll 	- memcpy.asm:188] 	memcpy
0x0130989a 	[libglesv2.dll 	- vertexdatamanager.cpp:164] 	gl::VertexDataManager::preRenderValidate(int,int,gl::TranslatedAttribute *)
0x01304f66 	[libglesv2.dll 	- context.cpp:1996] 	gl::Context::applyVertexBuffer(unsigned int,int,int,bool *,gl::TranslatedIndexData *)
0x013061a7 	[libglesv2.dll 	- context.cpp:2648] 	gl::Context::drawArrays(unsigned int,int,int)
0x012f7721 	[libglesv2.dll 	- libglesv2.cpp:1741] 	glDrawArrays
0x01c54f1e 	[chrome.dll 	- gles2_cmd_decoder.cc:3179] 	gpu::gles2::GLES2DecoderImpl::DoDrawArrays(unsigned int,int,int)
0x01c59122 	[chrome.dll 	- gles2_cmd_decoder_autogen.h:640] 	gpu::gles2::GLES2DecoderImpl::HandleDrawArrays(unsigned int,gpu::gles2::DrawArrays const &)

Review URL: http://codereview.appspot.com/3043042

git-svn-id: https://angleproject.googlecode.com/svn/trunk@480 736b8ea6-26fd-11df-bfd4-992fa37f6226

Only initialize through a temporary variable if the same symbol name is used in the initialization expression.
TRAC #13627
The previous patch can generate a lot of unnecessary temporary variables. By first checking whether the same symbol name is reused the clutter is reduced to an absolute minimum (typical shaders won't rely on this odd GLSL semantic behavior so the workaround is hardly ever needed).
Signed-off-by: Daniel Koch

Author:    Nicolas Capens

git-svn-id: https://angleproject.googlecode.com/svn/trunk@479 736b8ea6-26fd-11df-bfd4-992fa37f6226

TRAC #13627
GLSL allows to write things like "float x = x;" where a new variable x is defined and the value of an existing variable x is assigned. HLSL uses C semantics (the new variable is created before the assignment is evaluated), so we need to convert this to "float t = x, x = t;".
Signed-off-by: Daniel Koch

Author:    Nicolas Capens

git-svn-id: https://angleproject.googlecode.com/svn/trunk@478 736b8ea6-26fd-11df-bfd4-992fa37f6226

BUG=48
Review URL: http://codereview.appspot.com/3005042

git-svn-id: https://angleproject.googlecode.com/svn/trunk@476 736b8ea6-26fd-11df-bfd4-992fa37f6226

- lexer and parser reentrant
- line number handling automatic

Caveats:
- The preprocessor is still not thread-safe and full of bugs. I have another not-yet-ready patch to replace the preprocessor.
- The grammar files use options that are not supported by the old versions of flex and bison checked into compiler/tools. So I need to check-in the generated lexer-parser along with a shell script to generate them.

Review URL: http://codereview.appspot.com/2992041

git-svn-id: https://angleproject.googlecode.com/svn/trunk@475 736b8ea6-26fd-11df-bfd4-992fa37f6226

TRAC #14155
Issue=70
Signed-off-by: Daniel Koch

Author:    Nicolas Capens

git-svn-id: https://angleproject.googlecode.com/svn/trunk@474 736b8ea6-26fd-11df-bfd4-992fa37f6226

semantic for x and y coordinates rather than previous per-vertex
emulation when running on Shader Model 3.

Verified fix in Chromium with test case from bug.

BUG=http://code.google.com/p/angleproject/issues/detail?id=71
TEST=test case from bug
TBR=nicolas (LGTM'd by dgkoch)

Review URL: http://codereview.appspot.com/2921041

git-svn-id: https://angleproject.googlecode.com/svn/trunk@473 736b8ea6-26fd-11df-bfd4-992fa37f6226

TRAC #14054
The SetRenderTarget calls used in Blit::boxFilter() and Surface::swap() implicitly reset the scissor and viewport rectangles. So we need to ensure that the original rectangles get captured, and restored afterwards (the Context only keeps track of explicitly changed state).
Signed-off-by: Daniel Koch

git-svn-id: https://angleproject.googlecode.com/svn/trunk@472 736b8ea6-26fd-11df-bfd4-992fa37f6226

I tested that the samples still work and that they fail as expected, but do not crash, if the attribute list argument is null.

Review URL: http://codereview.appspot.com/2725042

git-svn-id: https://angleproject.googlecode.com/svn/trunk@471 736b8ea6-26fd-11df-bfd4-992fa37f6226

TRAC #14055
Signed-off-by: Daniel Koch

Author:    Nicolas Capens

git-svn-id: https://angleproject.googlecode.com/svn/trunk@470 736b8ea6-26fd-11df-bfd4-992fa37f6226

Fixed infinite loops and buffer overflow in byte_scan when scanning for integers and floating-point numbers.
- The byte_scan and associated functions are not very well written. I tried to clean them as much as possible without re-writing the whole thing.
- Replaced lBuildFloatValue function with atof. lBuildFloatValue was returning incorrect value anyway. The only reason it was working so far because we never used that value.
BUG=59623(crbug.com), 603333(bugzilla.mozilla.org)
Review URL: http://codereview.appspot.com/2655042

git-svn-id: https://angleproject.googlecode.com/svn/trunk@469 736b8ea6-26fd-11df-bfd4-992fa37f6226

I believe we don't support version 1 since we are missing things like the fixed function pipeline.

This is motivated by consistency with other EGL implementations, which return a v1 context if the EGL_CONTEXT_CLIENT_VERSION attribute is not specified.


Review URL: http://codereview.appspot.com/2760041

git-svn-id: https://angleproject.googlecode.com/svn/trunk@468 736b8ea6-26fd-11df-bfd4-992fa37f6226

Trac #14114
Reported in Bug=69

Signed-off-by: Nicolas Capens

git-svn-id: https://angleproject.googlecode.com/svn/trunk@467 736b8ea6-26fd-11df-bfd4-992fa37f6226

BUG=66
Review URL: http://codereview.appspot.com/2619041

git-svn-id: https://angleproject.googlecode.com/svn/trunk@466 736b8ea6-26fd-11df-bfd4-992fa37f6226

- Memory leaks
- Made compiling using multiple strings default. This was not getting exercised.
- Removed redundant copies of file data
- Handled empty files properly which were getting ignored
BUG=66
Review URL: http://codereview.appspot.com/2579041

git-svn-id: https://angleproject.googlecode.com/svn/trunk@465 736b8ea6-26fd-11df-bfd4-992fa37f6226

Fixed stack overflow in CPPextension function. An arbitrary size buffer was being used for extension name. Changed it to use MAX_SYMBOL_NAME_LEN.
- Also formalized the values for MAX_SYMBOL_NAME_LEN and MAX_STRING_LEN. They were being used as if there was a confusion whether it included the NULL terminator or not.
- Fixed some minor issues with code releated to the usage of MAX_SYMBOL_NAME_LEN and MAX_STRING_LEN.
BUG=59625 (crbug.com)
Review URL: http://codereview.appspot.com/2585042

git-svn-id: https://angleproject.googlecode.com/svn/trunk@464 736b8ea6-26fd-11df-bfd4-992fa37f6226

TRAC #13792
Signed-off-by: Daniel Koch

Author:    Nicolas Capens

git-svn-id: https://angleproject.googlecode.com/svn/trunk@463 736b8ea6-26fd-11df-bfd4-992fa37f6226

TRAC #13792
Signed-off-by: Daniel Koch

Author:    Nicolas Capens

git-svn-id: https://angleproject.googlecode.com/svn/trunk@462 736b8ea6-26fd-11df-bfd4-992fa37f6226

TRAC #13792
Signed-off-by: Daniel Koch

Author:    Nicolas Capens

git-svn-id: https://angleproject.googlecode.com/svn/trunk@461 736b8ea6-26fd-11df-bfd4-992fa37f6226

TRAC #13792
Signed-off-by: Daniel Koch

Author:    Nicolas Capens

git-svn-id: https://angleproject.googlecode.com/svn/trunk@460 736b8ea6-26fd-11df-bfd4-992fa37f6226

TRAC #13792
Signed-off-by: Daniel Koch

Author:    Nicolas Capens

git-svn-id: https://angleproject.googlecode.com/svn/trunk@459 736b8ea6-26fd-11df-bfd4-992fa37f6226

TRAC #13792
Signed-off-by: Daniel Koch

Author:    Nicolas Capens

git-svn-id: https://angleproject.googlecode.com/svn/trunk@458 736b8ea6-26fd-11df-bfd4-992fa37f6226

TRAC #13792
Signed-off-by: Daniel Koch

Author:    Nicolas Capens

git-svn-id: https://angleproject.googlecode.com/svn/trunk@457 736b8ea6-26fd-11df-bfd4-992fa37f6226

TRAC #13792
Signed-off-by: Daniel Koch

Author:    Nicolas Capens

git-svn-id: https://angleproject.googlecode.com/svn/trunk@456 736b8ea6-26fd-11df-bfd4-992fa37f6226

TRAC #13792
Signed-off-by: Daniel Koch

Author:    Nicolas Capens

git-svn-id: https://angleproject.googlecode.com/svn/trunk@455 736b8ea6-26fd-11df-bfd4-992fa37f6226

Added NSPR support for thread-local storage. Patch submitted by vladimirv. I have slightly modified the patch to still define ANGLE_OS_WIN and ANGLE_OS_POSIX.
BUG=54
Review URL: http://codereview.appspot.com/2497041

git-svn-id: https://angleproject.googlecode.com/svn/trunk@454 736b8ea6-26fd-11df-bfd4-992fa37f6226